xCloudinfo/Qwen3-VL-32B-Abliterated-xCloud
Abliterated vs Qwen/Qwen3-VL-32B-Instruct. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 14 of 64 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, Norms 0/373, MLP 14/316, Attention 14/364, lm_head 0/1
2 derivatives checked
- mradermacher/Qwen3-VL-32B-Abliterated-xCloud-i1-GGUF: Quantization (fewer layers)
- mradermacher/Qwen3-VL-32B-Abliterated-xCloud-GGUF: Quantization (fewer layers)